See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/382274558

Exploring Machine Learning Applications for Additive Manufacturing

Article · July 2024

CITATIONS READS

0 48

5 authors, including:

Kshitij Chouhan Dr Sumit Gupta

Amity University Amity University
2 PUBLICATIONS 0 CITATIONS 140 PUBLICATIONS 1,375 CITATIONS

SEE PROFILE SEE PROFILE

Sundeep Kumar
Malaviya National Institute of Technology Jaipur
49 PUBLICATIONS 184 CITATIONS

SEE PROFILE

All content following this page was uploaded by Sundeep Kumar on 16 July 2024.

The user has requested enhancement of the downloaded file.

Exploring Machine Learning
Applications for Additive Manufacturing

Kshitij Chouhan, Sumit Gupta , Vijay Chaudhary, Pallav Gupta,

and Sundeep Kumar

Abstract The use of data and algorithms to open the ways that human learn, which
gradually increase the accuracy, is all considered under a branch of computer science
and machine learning known as machine learning (ML). This is a rapidly growing
field of data science majorly includes the use of trained algorithms and statistical
methods to make classifications, which uncovering basic insights of data mining
projects. AM is a relatively new technology that has the potential to drastically alter
the future of digital manufacturing. In recent years, machine learning (ML) has
attracted a lot of attention in AM because of its amazing performance in data tasks
including classification, regression, and clustering. One of the primary justifications
is this very fact. The complexity of production systems leads to a variety of problems
with regard to design principles, standardization, and quality control. There is a
growing need, however, for more sophisticated and superior products.

Keywords Additive manufacturing · Machine learning · Supervised learning ·

Unsupervised learning

1 Introduction

The process known as additive manufacturing builds three-dimensional items by

superimposing successive layers of material in accordance with computer-aided
design (CAD) models [1]. The advancement of material science has made it possible
to create new types of substances that have exceptional functional qualities. It is
anticipated that these technological developments will usher in a new era of energy
harvesting, energy storage, the mechanical characteristics of components, and the
realization of self-powered materials [2]. The development of additive manufacturing

K. Chouhan · S. Gupta (B) · V. Chaudhary · P. Gupta

Department of Mechanical Engineering, ASET, Amity University Uttar Pradesh, Noida 201313,
India
e-mail: sgupta20@amity.edu
S. Kumar
Management Studies, Engineering College Ajmer, Rajasthan 305025, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 645
A. K. Shukla et al. (eds.), Recent Advances in Mechanical Engineering, Lecture Notes
in Mechanical Engineering, https://doi.org/10.1007/978-981-99-1894-2_55
646 K. Chouhan et al.

Fig. 1 Machine learning in AM

has made it feasible to create intricate geometric structures via the process of fabrica-
tion. The additive manufacturing process is now able to construct complicated pieces
using a variety of different kinds of materials [3]. Process of machine learning in
additive manufacturing is shown in Fig. 1.
The most recent developments in additive manufacturing are capable of producing
multi-scale, multi-material, and multi-functional objects, all of which are challenging
to construct using more conventional methods. There is still a gap between theoretical
design aspirations and real production capabilities, and this presents a challenge [4–
6]. Implementing algorithms that can identify defects and change printing settings in
real time is one way to get around these inherent hurdles and avert potential problems.
In order to automate analytical model creation and related operations, structural
modeling using ML may be trained on data from the original issue [7]. As a result of
recent advancements in machine learning, intelligent systems with human-like cogni-
tive ability are emerging, reshaping networked interactions on electronic market-
places and permeating our professional and personal life. Incorporating these tools
into businesses has helped boost productivity, employee engagement, and retention.
With the hope of assisting medical practitioners in making more effective use of ML
models, this study looks deeply into the current literature on automated machine
learning (AutoML) [2]. Only approximately 15% of hospitals routinely employ
machine learning in health care, and most of the health data obtained is never used
to construct prediction models that are successively incorporated into the clinical
setting [8].

2 Literature Review

Artificial Intelligence plays a vital role in healthcare sector in predicting the various
diseases. Machine learning via supervised and unsupervised learnings provides better
understanding of data to predict the various diseases. It is the process in which
machine is learning from data, which is provided to the machine in various forms
Exploring Machine Learning Applications for Additive Manufacturing 647

like image numerical data, etc. Machine uses its statistics for understanding the
pattern. Based on its observation, it predicts the outcome. This outcome is beneficial
for the people associated to it for taking future decision. To predict the outcome
machine use algorithms, modeling of data. In addition to this, machine learning has
the capability to handle large data, which comes from the previous records, reports,
digitally from the various electronical instruments [9]. Various techniques of machine
learning consist of total six techniques which are described below.

2.1 Support Vector Machine

This technique mainly deals with supervised learning model. Analysis of the provided
data consist of classification and implementation of regression analysis.

2.2 K-Nearest Neighbor

This technique mainly provides non-parametric classification of data. The output

depends on the cases of whether there is requirement for classification or regression.

2.3 Decision Tree

This technique mainly deals with supervised learning model. In this, the leaves and
branches symbolized class labels and conjunction of feature, respectively.

2.4 CART

Classification and Regression Tree Methodology is called as the CART. In this

technique, the target variable is represented as categorical for classification and
continuous for regression trees.

2.5 Fuzzy Logic

This technique is based on fuzzy theory in which the truth value of variable lies
between zero and one. The predictive modeling of machine learning is helpful for
healthcare sector [10]. There are various opportunities for machine learning like
648 K. Chouhan et al.

Fig. 2 Supervised learning

predicting the various risks of patients and informing the concerned authority for the
same.

2.6 Supervised and Unsupervised Learning

Supervised learning and unsupervised learning are two basic approaches to AI and
machine learning. While both use labeled data to help predict outcomes, one does
not. Both approaches have their advantages and disadvantages, but there are some
subtle differences between them.

2.6.1 Supervised Learning

In this, learning models are trained on labeled data and output is predicted based
on the features learned. Structured data are fed to machine learning models, where
machine is able to recognize the pattern between sample and extracted features. This
type of learning is useful to solve linear and nonlinear issues. Model of supervised
learning is shown in Fig. 2.
Various authors contributed in health care via implementing the supervised
learning. This direction implemented the novel method to examine the patient of
heart diseases by assessing and scoring the data collected by various sensors. Authors
utilized the electronic equipment like electrocardiogram sensors, SpO2 sensor for
examine the desired parameter of patient body. Advantage of this technology is that
it reduces the data complexity since its power consumption is less as compared with
other technology.

2.6.2 Unsupervised Learning

As its name suggests, these learnings do not require human intervention as the models
are trained using unstructured data samples. Machines operate with large dataset with
the help of program to reach for decision-making. Model of unsupervised learning
is shown in Fig. 3.
Unsupervised learning is more complex than supervised learning and required
much bigger dataset. However, it is quite useful and having a lot of application
than supervised learning. Non-labeled (unstructured) data are provided by human in
unsupervised learning via semi-supervised learning.
Exploring Machine Learning Applications for Additive Manufacturing 649

Fig. 3 Unsupervised learning

As its name suggest this type of learning method used for type of dataset, i.e.,
labeled as well as unlabeled dataset. The model is trained using few labeled dataset,
and result is predicted for unlabeled dataset.

2.6.3 Reinforcement Learning

In this type of learning, model uses the concept of reward and punishment [11]. When
the model predicts the correct result, it rewards itself otherwise gets punished. The
optimal solution in reinforcement learning is unknown to the/77system at the start
of the learning phase, so it must be determined iteratively. Sensible approaches are
rewarded in this process, while mistakes are punished. It is feasible that the system
will consider complicated environmental variables and react accordingly using this
approach. Consequently, the system develops its own answers through directing
incentives and punishments.

3 Applications of Machine Learning in Additive

Manufacturing

Additive manufacturing has applications in a variety of sectors and industries. Each

patient is unique in the medical and dentistry fields, so AM has a lot of potential
in customized and customized solutions. Personalized implants, medical models,
and saw guides are the most standard health clinical uses. Various applications of
machine leaning in additive manufacturing are shown in Fig. 4.

Fig. 4 Applications of AM
650 K. Chouhan et al.

4 Discussion

The current methods of additive manufacturing can make things with several scales,
multiple materials, and multiple functions, all of which are challenging to produce
using more conventional methods. There is still a mismatch between the expectations
of theoretical design and the capabilities of real production. This presents a challenge.
Recent academic efforts have been focused on finding solutions to these challenges
using more conventional optimization or simulation techniques. The ever-increasing
need for the production of components with intricate patterns has sparked a revolution
in the processes that are used in manufacturing. When it comes to the prototype of
designs that combine many functions and materials, additive manufacturing stands
out as a technology that has a lot of potential. Nevertheless, there are still obsta-
cles to overcome in the process of additive manufacturing. These obstacles include
mismatched material qualities, a lack of build consistency, and widespread flaws in
the printed object. By incorporating algorithms that can identify flaws and change
printing settings in real time, it is possible to circumvent the inherent difficulties that
are present.

5 Conclusion

According to the findings, directed energy deposition is only applicable to the creation
of implants, while sheet lamination is often reserved for the creation of medical phan-
toms or models. Powder bed fusion, material extrusion, and VAT photopolymeriza-
tion are now being employed for every category of production. Both binder jetting
and material jetting are not used in the production of tools, instruments, or compo-
nents for medical devices. Likewise, material jetting is not utilized in the production
of implants or biomanufacturing. The most often used materials include thermoplas-
tics, photopolymers, and various metals and metal alloys, such as titanium alloys.
Recent developments in binder jetting will, in the not-too-distant future, make previ-
ously unavailable alternatives available. It is proposed that in the future, a consistent
nomenclature can be used in the research that is carried out on the medical and
dental uses of additive manufacturing as well as the industrial side. A more method-
ical comparison of the ways in which various AM methods are used would be made
possible as a result of this.
 Exploring Machine Learning Applications for Additive Manufacturing 651

References

1. Gupta P, Jamwal A, Gupta S, Chaudhary V (2022) Modern approach towards additive manu-
facturing and 4d printing: emerging industries, challenges and future scope. Springer, Berlin,
pp 389–412
2. Mandache C (2019) Overview of non-destructive evaluation techniques for metal- based
additive manufacturing. Mater Sci Technol 35(9):1007–1015
3. Jha MK, Gupta S, Chaudhary V, Gupta P (2022) Material selection for biomedical application
in additive manufacturing using TOPSIS approach. Mater Today Proc
4. Talib S, Gupta S, Chaudhary V, Gupta P, Wahid MA (2021) Additive manufacturing: materials,
techniques and biomedical applications. Mater Today Proc 46: 6847–6851
5. Rao HS, Reddy DSK, Sharma C, Gupta S, Jamwal A, Agrawal R (2021) Assessment of key
barriers of sustainable additive manufacturing in Indian automotive company. In: Advances in
industrial and production engineering. Springer, Berlin, pp 245–253
6. Reddy K, Rao DSS, Sharma H, Gupta C, Shukla S, Kumar RK (2021) A: analysis of key enablers
of sustainable additive manufacturing: a case of Indian automotive company. In: Advances in
interdisciplinary engineering. Springer, Berlin, pp 281–290
7. Gupta S, Dangayach GS, Singh AK, Meena ML, Rao PN (2018) Implementation of sustainable
manufacturing practices in Indian manufacturing companies. Benchmarking Int J 25(7):2441–
2459
8. Weintraub WS, Fahed AC, Rumsfeld JS (2018) Translational medicine in the era of big data
and machine learning. Circ Res 123(11):1202–1204
9. Li R, Jin M, Paquit VC (2021) Geometrical defect detection for additive manufacturing with
machine learning models. Mater Des 206:109726–109726
10. Krishnan S, Gupta S, Kaliyan M, Kumar V, Garza-Reyes JA (2021) Assessing the key enablers
for Industry 4.0 adoption using MICMAC analysis: a case study. Int J Prod Performance Manag
11. Gentsch P (2018) AI in marketing, sales and service: How marketers without a data science
degree can use AI, big data and bots. Springer, Berlin

View publication stats